1. Field of the Invention
The present invention relates to an IP (Internet Protocol) network and an admission control method used therefor and, more particularly, to a transfer method at an optimum service level in an IP network.
2. Description of the Related Art
In recent years, IP networks have enjoyed their rapid growth to become global communication infrastructure having commerciality. With the growth, the IP networks are expected to serve as service infrastructure not only for conventional data communication but also for every other kind of communication application including telephone.
This expectation creates the need of transferring application traffic required to have different characteristics and service levels, that is, QoS (Quality of Service), at an optimum service level.
Currently proposed techniques which provide such QoS include Diffserv [Differentiated Service, IETF (Internet Engineering Task Force) RFC (Request For Comments) 2475] and MPLS (Multi Protocol Label Switching IETF RFC3031, IETF RFC2702).
Operation of a service-level-agreement-based Diffserv will be described with reference to FIG. 8. A core network manager terminal 23 of a core router 22 makes a service level agreement 511 (called SLA) with a manager terminal 25 of an access network 600.
The service level agreement 511 includes communication quality such as a band for use by traffic from the access network 600 with respect to each quality class and a delay provided by a core network 500, and a kind of traffic (application, user information) to be communicated in the quality class in question.
Based on the service level agreement 511, the core network manager terminal 23 conducts buffer assignment of edge routers 21a and 21b and the core router 22 in the core network 500 for each class and resource assignment 512 such as transfer scheduling setting. For providing service only to traffic designated by the agreement, the core network manager terminal 23 conducts publishing setting 514 at the edge router 21a.
Based on the traffic kind information indicated in the service level agreement 511, the edge router 21a classifies the inflow packet 514 and applies a DSCP (Diffserv Code Point) to the same as a class identifier. Then, the core router 22 within the domain conducts queuing and transfer scheduling for each class based on the value of the DSCP.
This enables communication quality to be provided for each class so as to meet properties of an application. At this time, the admission of the edge router 21a monitors an inflow of traffic to discard traffic 515 whose volume exceeds a contracted volume of traffic or transfer the same similarly to common traffic without receiving contracted service [called BE (Best Effort)].
On the other hand, in an MPLS, a label of a fixed length is applied to a packet which will be transferred based on the value of the label. By implicitly controlling an LSP (Label Switched Path) as a path on which the packet is transferred, an optimum path 513 can be provided based on a requested QoS of traffic.
Commonly used as a charging method is a system in which the access network manager terminal 25 pays an agreement fee for the service level agreement 511 made with the core network manager terminal 23, while collecting monthly-fixed charges from user terminals 24a and 24b. Another system is recently provided of setting a charge for each quality class used by the user terminals 24a and 24b. 
With the above-described conventional architecture, however, even using the same communication application, the user terminals 24a and 24b might not dare to have high quality communication at the cost of a heavy charge. It is therefore inappropriate to designate a certain application as a kind of traffic to uniformly assign a high-quality class to the traffic of the application and charge for the same.
In addition, since in many of communication applications used in the IP networks, the user terminals 24a and 24b autonomously start communication, the access network manager terminal 25 is not allowed to dynamically control to which user' application traffic assigned is contracted service. As a result, a volume of inflow traffic of each quality class to the core network 500 exceeds a contracted volume of traffic, so that contracted communication quality might not be guaranteed for the user terminal 24a. 
Moreover, when making the service level agreement 511 with the core network manager terminal 23 by paying a fee, the access network manager terminal 25 is not allowed to charge the user terminal 24a according to the use of high-quality communication service in order to collect the fee.
Furthermore, in a case where there exist a plurality of entrance edge routers through which the access network 600 connects to the core network 500, even if resource assignment including path control is conducted from the entrance edge routers to an exit edge router by an MPLS or the like, appropriate resource assignment is not possible because it is impossible to predict through which entrance edge router the application traffic flows.
A further system is provided in which when communicating, a user terminal makes a request for ensuring communication quality and a network conducts setting for ensuring the communication quality in response to the request. Setting for ensuring communication quality, however, includes calculation of an optimum path and a necessary amount of buffer, setting of a calculated path and buffer assignment setting of all the routers on the path, and accordingly costs time for setting.
Thus, if every time the user terminal conducts communication, it makes a request for ensuring communication quality and then the network conducts setting for ensuring the communication quality, it takes time before setting is completed to delay communication start at the user terminal.